Product-Manager-Skills: 77 PM Frameworks as Claude Code Skills for AI-Assisted Product Work

The problem this repository solves: why generic AI output fails PM work

The README frames the core problem directly: generic AI output is a product manager's worst enemy. When an agent is told to write a PRD without shared context, it produces a generic document that no stakeholder trusts and no engineer can act on. The agent guesses at structure, borrows from its training data, and fills in reasoning that is plausible but not grounded in the product's actual situation.

Product-Manager-Skills addresses this by giving both the PM and the agent the same professional foundation before the work starts. Each skill file includes the reasoning behind the framework, the failure modes to avoid, and the judgment calls a senior PM would make in applying it. The README describes the goal as dual: functional and pedagogic in equal measure. Skills equip agents to do PM work at a professional level, and they teach the human PM the reasoning behind each framework so they can explain it, adapt it, and pass it on.

The README explicitly states that neither function is a byproduct of the other. This design choice means reading a skill file teaches the PM something, and loading it into an agent session produces genuinely structured output rather than a filled-in template.

Repository layout: skills/, commands/, catalog/, and the Claude plugin

The repository is organized into functional directories. The skills/ directory holds the core frameworks, each under a subdirectory named after the skill with a SKILL.md file containing the full framework instructions. The commands/ directory holds agent command definitions. The catalog/ directory provides an index of available skills. The .claude-plugin/ directory contains the integration files for Claude Code's plugin system. The dist/ directory holds built artifacts, and the docs/ directory holds supporting documentation.

Navigating the repository starts with START_HERE.md at the root, which the top-level entries list confirms is present. The README organizes skills by activity category rather than alphabetically, so a PM working on a specific problem can navigate by task: strategy framing, stakeholder alignment, customer discovery, prioritization, deliverable writing, validation, finance and growth, market intelligence, or aging product management.

To use the repository with Claude Code, clone it and load the appropriate skills:

git clone https://github.com/deanpeters/Product-Manager-Skills

Each skill in skills/<name>/SKILL.md can be loaded into an agent session. The .claude-plugin/ directory supports registering the skills collection as a Claude Code plugin, which makes skills available without referencing individual file paths in each session.

The strategy and stakeholder skill group

The framing and strategy section covers three skills. The problem-framing-canvas implements MITRE's Look Inward / Look Outward / Reframe sequence, which the README describes as stopping teams from solving the wrong problem. The positioning-statement implements Geoffrey Moore's positioning template for defining who you serve, what problem you solve, and how you are different. The product-strategy-session covers the full strategy arc from positioning through problem framing, solution exploration, and roadmap in a two-to-four week facilitation structure.

The stakeholder alignment group adds three more skills. The stakeholder-identification skill maps every stakeholder before engaging anyone, running through a broad brainstorm, classifying stakeholders as allies, audiences, or influencers, applying R/P/D marking, and applying an equity lens before narrowing to priority targets. The stakeholder-mapping skill runs two complementary grids: Power times Interest for engagement strategy, and Impact times Power for whose voice to elevate, then compares the two to find gaps. The stakeholder-engagement-advisor goes further, producing per-stakeholder message framing, channel selection, cadence recommendations, and a named next action.

These skills are designed to be used in sequence. The strategy session consumes the positioning output. The stakeholder engagement plan builds on the stakeholder map. The README presents them as a structured flow rather than a menu of isolated tools.

Discovery, prioritization, and PM deliverable skills

The customer discovery group provides three skills organized as a complete discovery cycle. The discovery-interview-prep skill plans Mom Test-style interviews based on stated research goals. The opportunity-solution-tree skill generates opportunities and solutions, then recommends the best proof-of-concept to run first. The discovery-process skill covers the full cycle from framing through research, synthesis, and validation over three to four weeks.

The prioritization section offers three skills. The prioritization-advisor skill asks three to five diagnostic questions about the team's context, then recommends the right scoring framework: RICE, ICE, Kano, or an alternative. The epic-breakdown-advisor skill splits large epics using Richard Lawrence's nine splitting patterns. The roadmap-planning skill covers the full sequence from input gathering through epic definition, prioritization, sequencing, and communication in one to two weeks.

For writing PM deliverables, the user-story skill implements Mike Cohn's format with Gherkin acceptance criteria and explicit anti-patterns. The prd-development skill produces a structured PRD moving from problem definition through personas, solution, metrics, and user stories in two to four days. The press-release skill implements Amazon's Working Backwards approach, which requires clarifying the product vision before writing a line of specification. Each of these skills includes the reasoning behind the format, not just the template, so the agent can explain why a particular section is structured as it is.

The market intelligence chain: how four skills compose into a pipeline

The market intelligence section demonstrates how skills in this repository are designed to feed each other's outputs. The intel-discipline-advisor is the entry point: it triages the intelligence question on the PM's desk into the right disciplines and cadence, then identifies which executing skill to run next. This triage step prevents a common failure mode where the PM starts with a battle card template when the real gap is a market landscape assessment.

The investigation chain runs in sequence: market-landscape-scan produces a landscape snapshot, competitive-research-snapshot consumes that snapshot and adds product comparison data, competitive-intel-watch adds a monitoring cadence, and battle-card-builder consumes the prior skill's schema to produce competitive battle cards. The README describes this as turning research from a one-off deck into a repeatable cadence.

The tam-sam-som-calculator addresses market sizing in three distinct modes: using the PM's own numbers, running a guided interview to collect inputs, or performing autonomous bottom-up research that produces a result a skeptical CFO can attack one assumption at a time. The README specifies this last mode explicitly, which is a concrete constraint: the bottom-up research output must be structured well enough to withstand interrogation of individual assumptions rather than accepting the total as a single number.

The competitive-analysis-process skill provides the six-step umbrella that covers the full intelligence effort: landscape scan, product comparison, customer needs analysis, business baseline, positioning, and strategic direction. It serves as the orchestrating framework when the team needs to run all stages rather than a targeted investigation.

The lifecycle-play-advisor and the aging product workflow

One of the more specific skill groups covers products that have stopped growing. The lifecycle-play-advisor is the entry point: it asks seven transition questions to establish what lifecycle stage the product is in, then picks the play: extend, replace, or retire. The README notes explicitly that it will output "nothing yet" when the honest answer is that the product is not ready for a transition decision. This constraint prevents the agent from defaulting to a confident but premature recommendation.

The product-lifecycle-plays skill is the framework behind this call. It provides the PLC strategy grid and the seven reasoning steps the lifecycle-play-advisor uses. Having both the advisor and the underlying framework as separate skills means the PM can load just the advisor for an interactive session or load both to understand the full reasoning.

For growing products, the organic-growth-advisor skill implements the McKinsey Growth Pyramid triage: it diagnoses whether the constraint is in new segments, geographies, channels, or products. The business-health-diagnostic skill focuses on SaaS health specifically, examining growth, retention, efficiency, and capital using the PM's actual metrics. The feature-investment-advisor produces a build or do-not-build recommendation using revenue impact, cost, ROI, and strategic value as inputs.

Comparison with standalone PM prompt libraries and the license consideration

Standalone PM prompt libraries, such as collections of ChatGPT prompts for common product management tasks, represent the most common alternative approach. These libraries typically provide individual prompts meant to be pasted into a chat session. The difference in design is fundamental: a standalone prompt produces output in a single exchange, while a Product-Manager-Skills skill file encodes a multi-step framework that the agent executes over several turns, with each step depending on the outputs of prior steps.

The stakeholder-identification skill, for example, does not produce a stakeholder list in one pass. It runs through a broad brainstorm, applies classification criteria, introduces an equity lens, and then narrows to priority targets. Each stage informs the next. A standalone prompt for stakeholder analysis would produce a list based on the agent's training data; the SKILL.md file drives a structured process that uses the PM's specific context.

The license situation is worth noting for adoption decisions. The LICENSE file is present in the repository, but GitHub classifies it as NOASSERTION, meaning the license text was not recognized as a standard OSI-approved identifier. Teams that need to verify usage rights before adopting the repository should read the LICENSE file directly rather than relying on GitHub's classification. The repository has releases through v0.84, and the latest release was dated 2026-08-10.

Editorial conclusion

Product-Manager-Skills is the right tool for product managers who work with Claude Code, Codex, or other AI coding assistants and want their agent to apply specific PM methods rather than producing generic output. The skill files give the agent the reasoning behind each framework, not just a template, which means it can explain choices and adapt to context. It is not useful for teams that do not use Claude Code or a compatible AI assistant: the SKILL.md files are instructions for an agent, not standalone documents you hand to a stakeholder. Before adopting it, verify the license terms in the LICENSE file, since GitHub classifies it as NOASSERTION and it is not a recognized standard open-source license.

Frequently asked questions

Official sources

1. deanpeters/Product-Manager-Skills on GitHub
2. Issues
3. README
4. Releases